Bulletin of the Seismological Society of America, Vol. 96, No. 6, pp. 2449 2456, December 2006, doi: 10.1785/0120050152 Short Note Frequency-Dependent Lg Attenuation in the Indian Platform by S. Mitra,* K. Priestley, V. K. Gaur, and S. S. Rai Abstract We use seismograms from regional earthquakes recorded on digital seismographs in peninsular India to determine the frequency-dependent Q of Lg for the Indian platform. We measure Lg attenuation by determining the decay of spectral amplitudes with distance. The available data suggest some spatial variation in attenuation but a much denser ray-path coverage would be required to validate such observations. We, therefore, combine all the measurements of overlapping regions that span both the shield and intervening terranes to obtain an average value of attenuation for the Indian platform: Lg Q 665 10 with the frequency exponent n 0.67 0.03. This average value of Lg attenuation for the Indian platform is similar to the average for other stable regions of the globe. Introduction Lg waves are short-period, higher-mode surface waves that propagate in the continental crust as a coherent package of energy with typical group velocities of 3.5 km sec 1 (Press and Ewing, 1952; Båth, 1954). For continental earthquakes at regional distances the Lg phase is often a largeamplitude arrival typically constituting the main portion of the wave-train energy. High-frequency, higher-mode Rayleigh waves combine to produce the vertical component Lg phase (Knopoff et al., 1973). Lg can also be considered to arise from the interference of multiple reflected S waves propagating in the crustal wave guide (Bouchon, 1982; Kennett, 1985). Because virtually all the energy contributing to the Lg wave is in the form of shear-wave energy trapped in the crust, the Lg amplitude is strongly dependent on the crustal structure and therefore provides discriminating information about the average crustal shear-wave velocity and attenuation between the source and receiver. The Lg wave train formed by the superposition of the different S multiples not only builds to a clear amplitude maximum but often has a relatively long coda that may persist for several minutes after the onset of Lg. Lg coda measured over a long part of the seismogram results in a rather effective averaging of the scattered field, thereby acquiring a remarkably stable envelope magnitude that proves quite attractive for the estimation of earthquake magnitude (Mayeda and Walter, 1996) from fewer station data. Because of their well-constrained domain of propagation and strong stable amplitudes over a large epicentral distance, both Lg and Lg coda wave trains are well suited for attenuation measurements. In addition, spatial de- *Present address: Department of Geology and Geophysics, Indian Institute of Technology, Kharagpur 721302, India. cay rates of their spectral amplitudes provide insights in the assessment of earthquake hazards. In this work we determine the frequency-dependent Q of the Indian platform from the spatial decay of the spectral amplitudes of Lg wave trains since they were stronger on our records than the Lg coda. Little is known about the propagation characteristics of Lg in India. Because the Indian shield has low seismicity and until recently digital broadband seismic data were only available for the GEOSCOPE station at Hyderabad (HYB) (Fig. 1), few attempts at measuring Lg attenuation of the various Indian regions have been made. Singh et al. (2004) used data for four earthquakes to measure the frequency-dependent Q of Lg waves for the Indian shield region. Mandal and Rastogi (1998) measured the frequency dependence of coda Q c (i.e., the total quality factor of the medium) for the Koyna Warna seismic zone close to the west-central coast of India. Our study covers most of the Indian platform and is based on better ray-path coverage than that used in previous studies. The Indian shield is a stable region composed of three distinct Archean cratons: the Dharwar craton, the Singhbhum craton, and the Aravali craton (Fig. 1), which have remained a coherent unit since the Late Archean or early Proterozoic (Naqvi and Rogers, 1987). This coherence of the Indian Shield and its nearuniform crustal thickness of 35 40 km (Gaur and Priestley, 1997; Rai et al., 2003; Gupta et al., 2003) provide the rationale for seeking an average characteristic Lg Q-value for the entire Indian platform. Data and Analysis Data for this study consist of regional, digital seismograms of three earthquakes: (1) Jabalpur in central India, 2449
2450 Short Note Figure 1. Geological map of the Indian subcontinent, taken from Naqvi and Rogers (1987), showing the location of earthquakes (denoted by the filled circles), the broadband seismic stations (denoted by the filled squares) that recorded data for this study and the ray paths between them. (2) Chamoli in the Himalaya in northern India, and (3) Bhuj in western India. Source information for these earthquakes taken from the Preliminary Determination of Epicenter (PDE) catalog is given in Table 1. Seismograms were recorded at 17 broadband seismograph stations in peninsular India (Table 2), operated by the Cambridge University, Indian Institute of Astrophysics (CU-IIA), the National Geophysical Research Institute (NGRI), and the Indian Meteorological Department (IMD). Figure 1 shows the location of the three earthquakes and the seismic stations. The source receiver travel paths used in this study provided a good sampling of the crust of the central and south Indian Shield (Fig. 1). We measure Lg attenuation by determining the decay of spectral amplitudes with distance. For shallow earthquakes, the S wave is the dominant phase on seismograms at distances less than about 100 km, whereas Lg dominates at dis-
Short Note 2451 Table 1 Source Information (PDE Catalog) of the Three Earthquakes Used in the Lg Q Measurements No. Origin Date Time (hh:mm:ss) Latitude Longitude Depth (km) Magnitude 1 21 May 1997 22:51:30 23.104 80.118 39 6.0 2 28 March 1999 19:05:11 30.512 79.403 15 6.4 3 26 January 2001 03:16:42 23.400 70.280 20 7.5 Table 2 Location and Seismometer Type of NGRI, IMD, and CU-IIA Stations That Recorded Data Used in This Study Network Station Latitude ( N) Longitude ( E) Elevation (m) Seismometer Type NG BGL 13.021 77.570 791 CMG-3ESP NG GBA 13.564 77.357 681 CMG-3ESP NG SLM 16.101 78.894 368 CMG-3ESP SE TRV 8.510 76.960 64 STS-2 SE MNG 12.941 74.822 11 CMG-40T SE MDR 13.070 80.250 15 STS-2 SE PUNE 18.530 73.850 560 STS-2 SE NGP 21.101 79.062 304 CMG-40T SE JHN 25.465 78.539 250 CMG-40T SE KARD 17.307 74.183 561 STS-2 SE VISK 17.721 83.328 10 STS-2 SE BLSP 22.129 82.131 271 STS-2 SE BHPL 23.241 77.424 502 STS-2 SE BHUJ 23.254 69.654 101 STS-2 SE BOKR 23.794 85.885 307 STS-2 SE AJMR 26.479 74.643 492 STS-2 CU NND 19.107 77.287 314 CMG-3T tances greater than about 100 km (Singh and Herrmann, 1983). The distance range used in this analysis (Table 3) has a minimum of 233 km, providing clearly observed Lg phases. Lg on the vertical component is extracted using a velocity window of 3.6 2.8 km sec 1. Record sections of verticalcomponent seismograms for the three events are shown in Figure 2, with Lg windows used for analysis marked on each seismogram. To determine the Lg attenuation, instrument-corrected, displacement-amplitude spectra were computed for the vertical component Lg phase. Noise levels in the data were determined by processing in a similar manner an equal length time window before the initial P-wave arrival time. In an attempt to eliminate random errors in the amplitude spectra, only those frequencies that had a signal-to-noise ratio greater than two were considered in the analysis. The Lg spectra were inverted for Q(f) using S(f) pfr/vq(f) A(R, f) e (1) G(R) where, A(R, f) is the spectral amplitude observed at a distance R and frequency f, S(f) is the source term, G(R) isthe geometrical spreading term, v is the average group velocity, Table 3 Number of Stations That Recorded Each Earthquake, the Total Distance Range Covered for the Lg Q Measurements, Q 0, and Frequency Dependence (n) for the Three Earthquakes Earthquake No. No. of Stations Distance Range (km) Q 0 Value Frequency Dependence (n) 1 9 233 1071 521 21 0.74 0.10 2 10 565 2001 663 10 0.74 0.03 3 7 654 1795 869 45 0.64 0.11 taken as 3.5 km/sec for Lg. This model does not take into account scattering or radiation pattern effects and may therefore yield an apparent rather than the intrinsic Q. The effect of the source radiation pattern is minimized because the Lg phase is constructed as a superposition of many higher-mode surface waves (Knopoff et al., 1973; Panza and Calcagnille, 1975) sampling a major portion of the focal sphere. Lg is modeled accordingly as constructively interfering highermode surface waves (Knopoff et al., 1973; Nuttli, 1973; Panza and Calcagnille, 1975) and it is therefore assumed that the frequency-domain geometrical spreading scales with the square root of distance ( R). Taking log 10 of equation (1) gives pf log10 e log10 A 0.5 log10 R log10 S R (2) vq This is the equation of a straight line, whose intercept is given by the source term and slope by the Q term. For each earthquake, we plot (log 10 A 0.5 log 10 R) versus R and perform a linear regression to determine Q at each frequency. Since Lg is an Airy phase (Nuttli, 1973), we are justified in using a frequency-independent travel time and hence a speed (v) of 3.5 km sec 1 in (2). Writing the frequency-dependent Q in the form n Q(f) Qo f, (3) where Q o is the Q at f 1 Hz, and n gives the frequency dependence of the quality factor, we can rewrite (3) as log Q log Q n log f. (4) 10 10 o 10 This again is the equation of a straight line and a linear regression is accordingly performed over log 10 Q versus log 10 f, to determine values of Q o and n, from the intercept and slope of the regressed line, respectively. In Figure 3 we plot log 10 (A) 0.5 log 10 R versus R for all earthquakes at frequencies 0.6, 0.75, 1, 1.2, 1.5, 2, 3, 4, 5, and 6 Hz. The slope of the regression line in Figure 3 when substituted in (2) yields the Q-value at these frequencies. A comparison of these plots shows that for the lower frequencies, Q is systematically higher for the paths sampled by the Bhuj earthquake. Figure 4 shows the path coverage
2452 Short Note Figure 2. Record section plot of the vertical component seismograms of the three earthquakes used for the Lg analysis: (a) 21 May 1997 (Jabalpur); (b) 28 March 1999 (Chamoli); and (c) 26 January 2001 (Bhuj). The velocity window from 3.6 to 2.8 km/ sec is denoted by the pair of arrows above or below the seismograms. for the individual earthquakes and plots of log 10 Q versus log 10 f. The regression line through each dataset gives a value for Q 0 and n in Table 3 and Figure 4. Finally, we perform a weighted fit of the data from all three earthquakes to obtain an average value of Q 0 and n for the Indian Shield region (Fig. 5) 0.67( 0.03) Q0 655( 10) f 0.6 f 6.0 Hz. (5) This is obtained by fitting the average of the earthquakederived Q-functions, weighted by the standard deviation at each frequency. Results and Discussion We determine Lg attenuation, using digital records at 17 broadband seismic stations, of three regional earthquakes: one in the central Himalayan belt and the other two in the western and central part of the Indian shield. We selectively used only the high signal-to-noise (S/N) records which is
Short Note 2453 Figure 3. Plot of the log 10 A 0.5 log 10 R versus (R) for frequencies between 0.6 and 6 Hz for the three earthquakes.
2454 Short Note Figure 4. (a) Result for the 21 May 1997 Jabalpur earthquake in central India. (Left) Ray-path coverage map. (Right) Plot of Log(Q) versus Log(f). The error bars are plotted over each data point. (b) Result for the 28 March 1999 Chamoli earthquake in the Himalaya. (c) Result for the 26 January 2001 Bhuj earthquake in western India.
Short Note 2455 Figure 5. Combined result for the three earthquakes. (Left) Ray-path coverage for the Indian Shield. (Right) Plot of the weighted average of the Log(Q) versus Log(f) values. reflected in the fairly well-resolved log(a) plots shown in Figure 3. We find that within the Indian platform region, the Lg Q is clearly frequency dependent with the frequency exponent n 0.67 0.03. Our results for Lg Q are better constrained for paths joining the Himalaya and sites in the south Indian shield (Fig. 4b) that include a substantial part of the stable cratons. These yield the expected high Lg Q value of 663 10, and n 0.74 0.03. For paths spanning central India (Fig. 4a), which include both cratons and intervening terranes of more variable ages, the values are somewhat lower: Lg Q 521 21 and n 0.74 0.1. However, the Q-value is substantially higher and has larger scatter for paths (Fig. 4c) that include the west coast of India: Lg Q 869 45, and n 0.64 0.11. The latter high Lg Qvalue is therefore largely weighted by the Western Ghats that constitute a narrow ( 50 km) moderate relief ( 1500 m) ridge which hugs the western coast most of the way and could possibly be the result of preferential energy channeling along the Ghats. This value is also higher than that determined by Mandal and Rastogi (1998) for a small region of 50-km diameter array around Koyna on the west coast. But this can be explained because their results are valid only for a near-surface laterally limited volume of rocks subject to frequent reservoir-induced seismicity (Langston, 1976), and are therefore not directly comparable with our results corresponding to continental scale paths. The north-central Indian shield and the south Indian shield are characterized by significantly different values of Q. From a knowledge of recent earthquakes in India we know that the south Indian Shield region is relatively more stable and shows low seismic activity compared with north and central India where there have been several recent, moderately large earthquakes: Uttar-Kashi (1992), Latur (1993), Jabalpur (1997), Chamoli (1999), and Bhuj (2001). However, available data allow only a coarse regionalization and a much denser ray-path coverage would be required to validate such observations. We, therefore, combine all the previous three measurements of overlapping regions that span both the shield and intervening terranes to obtain an average value of attenuation for the Indian platform: Lg Q 665 10 and n 0.67 0.03. Our average Lg Q values for the Indian platform are lower than the value of 800 determined by Singh et al. (2004) for the Indian shield using four regional events, three of which were common to our study. Their frequency exponent of 0.42 is also more than 50% lower than our average value of 0.67. Shi et al. (1996) found Lg Q-values for several subregions of northeastern United States to vary from 905 for the Adirondack Mountains with exposed Precambrian Grenville basement, which could be an analog of the Indian shield, to 561 for the Appalachian plateau and folded zone and a frequency exponent of 0.4 to 0.47. Baqer and Mitchell (1998) studied a larger region of the eastern United States extending from the Rocky Mountains to the Atlantic coast and obtained Lg Q-values of 450 750. However, our average Q-value for the Indian platform is similar to the average for other stable regions of the globe: southern Africa (Congo and Kalahari) cratons, the West African craton (Xie and Mitchell, 1990), and the Siberian craton (Q 400 600) (Mitchell, 1995). Most notably our Lg Q as well as frequency dependence are quite similar to the new continent wide maps of Lg coda Q and frequency dependence in Eurasia (Brian Mitchell, pers. comm., 2005). In particular, the similarity between our Lg Q and Lg coda values underline the implication that attenuation across India is dominated by intrinsic, rather than scattering mechanisms.
2456 Short Note Acknowledgments S.M. thanks Cambridge Commonwealth Trust for providing a scholarship to study at the University of Cambridge where this work formed a portion of his Ph.D. research. This project was supported in part by a grant from the Department of Science and Technology, Government of India, and by Bullard Laboratories of the University of Cambridge. Considerable logistic support was provided by the director of the National Geophysical Research Institute at Hyderabad. V.K.G. thanks the Directors of IIA and C-MMACS for providing infrastructure support. The manuscript has benefited from the review and comments of Keith Koper (associate editor) and Brian Mitchell. The maps shown in Figures 1, 4, and 5 were made with Generic Mapping Tools (Wessel and Smith, 1995). This is Cambridge University Department of Earth Sciences contribution ES8483. References Baqer, S., and B. J. Mitchell (1998). Regional variation of Lg coda Q in the continental United States and its relation to crustal structure and evolution, Pure Appl. Geophys. 153, 613 638. Båth, M., (1954). The elastic waves Lg and R g along Euroasiatic paths, Arkiv. Geofysik. 2, 295 342. Bouchon, M. (1982). The complete synthesis of seismic crustal phases at regional distances, J. Geophys. Res. 87, 1735 1741. Gaur, V. K., and K. F. Priestley (1997). Shear wave velocity structure beneath the Archaean granites around Hyderabad, inferred from receiver function analysis, in Proceedings of the Indian Academy of Sciences, Earth Planet. Sci. 106, 1 8. Gupta, S., S. S. Rai, K. S. Prakasham, D. Srinagesh, B. K. Bansal, R. K. Chadha, K. Priestley, and V. G. Gaur (2003). The nature of South Indian crust: Implications for Precambrian crustal evolution, Geophys. Res. Lett. 30, no. 1, 1 4. Kennett, B. L. N. (1985). On regional-s, Bull. Seism. Soc. Am. 75, 1077 1086. Knopoff, L., F. Schwab, and E. Kausel (1973). Interpretation of Lg, Geophys. J. R. Astr. Soc. 33, 389 404. Langston, C. A. (1976). A body wave inversion of the Koyna, India, earthquake of December 10, 1967, and some implications for body wave focal mechanisms, J. Geophys. Res. 81, 2517 2529. Mandal, P., and B. K. Rastogi (1998). A frequency-dependent relation of coda Q for Koyna-Warna region, India, Pure Appl. Geophys. 153, 163 177. Mayeda, K., and W. Walter (1996). Moment energy, stress drop, and source spectra of western United States earthquakes from regional coda envelope, J. Geophys. Res. 101, 11,195 11,208. Mitchell, B. J. (1995). Anelastic structure and evolution of the continental crust and upper mantle from seismic surface wave attenuation, Rev. Geophys. 33, no. 4, 441 462. Naqvi, S. M., and J. J. W. Rogers (1987). The Precambrian Geology of India, Oxford University Press, New York. Nuttli, O. W. (1973). Seismic wave attenuation and magnitude relations for eastern North America, J. Geophys. Res. 78, 876 885. Panza, G. F., and G. Calcagnille (1975). Lg, Li and R g from Rayleigh modes, Geophys. J. R. Astr. Soc. 40, 475 487. Press, F., and M. Ewing (1952). Two slow surface waves across North America, Bull. Seism. Soc. Am. 42, 219 228. Rai, S. S., K. Priestley, K. Suryaprakasam, D. Srinagesh, V. K. Gaur, and Z. Du (2003). Crustal shear velocity structure of the south Indian shield, J. Geophys. Res. 108, no. B2, doi 10.1029/2002JB001776. Shi, J. K., W-Y. Kim, and P. G. Richards (1996). Variability of crustal attenuation in the northern United States from Lg waves, J. Geophys. Res. 101, 25,231 25,242. Singh, S. K., and R. B. Herrmann (1983). Regionalization of crustal coda Q in the continental United States, J. Geophys. Res. 88, 527 538. Singh, S. K., D. Garcia, J. F. Pacheco, R. Valenzuela, B. K. Bansal, and R. S. Dattatrayam (2004). Q of the indian shield, Bull. Seism. Soc. Am. 94, 1564 1570. Wessel, P., and W. H. F. Smith (1995). New version of the generic mapping tool release, EOS Trans. AGU 76, 329. Xie, J. K., and B. J. Mitchell (1990). A back-projection method for imaging large-scale lateral variation of Lg coda Q with application to continental Africa, Geophys. J. Int. 100, 161 181. Bullard Laboratories University of Cambridge Cambridge CB3 0EZ, United Kingdom (S.M., K.P.) Indian Institute of Astrophysics Bangalore 560034, India (V.K.G.) Center for Mathematical Modelling and Computer Simulations Bangalore 560034, India (V.K.G.) National Geophysical Research Institute Hyderabad 500007, India (S.S.R.) Manuscript received 27 July 2005.